high performance computing on graphics processing units: hgpu.org

Handwriting recognition is having high demand in commercial & academics. In recent years lots of good work has been done on hand written digit recognition to improve accuracy. Handwritten digit recognition system needs larger dataset and long training time to improve accuracy & reduce error rate. Training of Neural Networks for large data sets is very time consuming task on CPU. Hence, in this paper we presented fast efficient artificial neural network for handwritten digit recognition on GPU to reduce training time. Standard back propagation (BP) learning algorithm with multilayer perceptron (MLP) classification is chosen for this task & implemented on GPU for parallel training. This paper focused on specific parallelization environment Compute Unified Device Architecture (CUDA) on a GPU hence effectively speedup training & reduce training time.